Toward high level data fusion for conflict resolution
Zeinab Nakhaei, Ali Ahmadi · 2017
Conflict resolution is the problem of finding true value among different and controversial facts about a single entity provided by different data sources such as web sites. All of the previous studies have relied on estimating two basic parameters, accuracy of data and reliability of sources. These methods are dealing with some challenges such as knowing distribution of data a priori or assumption about dependency of sources which are due to low level fusion. In lower levels of data fusion we usually engage with details of data generation process, while in higher levels of abstraction the relationship between objects becomes more important. In this paper, we propose the approach of High Level Conflict Resolution (HLCR) based on graphical model for conflict resolution which performs in high level data fusion and uses the relations between objects for inferring truth value. We have specified the well-known JDL model for the problem of conflict resolution. Evaluation results showed that our proposed approach outperforms existing conflict resolution techniques especially where the number of reliable sources is low.